More than 40 million family members provide informal care to older individuals annually, making them an essential component of the health care system in the US. Almost 11 million Americans provide unpaid care for a person with Alzheimer's disease or other dementia. Health care professionals typically rely heavily on self-report by family members to understand the dynamics of the caregiving dyad, the quality of caregiving, and the impact of dementia on everyday life. Though invaluable, this information may be devoid of context, or biased due to caregiver reticence or imperfect recall. First-person vision (FPV) technology offers a novel approach to understand the challenges of caregiver/care-recipient interactions, and caregivers' responsiveness to clinicians' suggestions for handling difficult caregiving challenges. Equipped with tiny sensors and worn on eyeglasses or clothing, FPV devices capture video and sound within the user's field of view. These devices can document actual interaction, permit identification of antecedent circumstances or consequences, and provide the basis for targeted intervention designed to assure safe interactions and to prevent, eliminate, or ameliorate distressing care-recipient behaviors. This project with caregiving dyads aims: to describe the feasibility, usability, and interpretability of FPV device use at home (Aim 1); to explore the effect of nurse- delivered intervention, informed by FPV data, on reducing problem behaviors and caregiver burden when delivered in person within 2 weeks (Aim 2) or delivered by phone in near real-time, that is, within 1 hour, of problem behavior events (Aim 3); and to develop algorithms for an automated system capable of generating real-time Machine Learning responses that promote effective handling of problem behaviors (Aim 4). This project will involve 40 caregiving dyads comprising family caregivers age 21 or older who experience these problem behaviors and care-recipients age 50 or older with moderate to severe dementia (MMSE score <18) due to Alzheimer's disease, frontal temporal dementia, Lewy body dementia, or vascular dementia. Baseline data will include demographic, health, and caregiving information including problem behaviors, functional status, and attitudes toward technology. Caregivers among the first 20 dyads will wear two FPV devices during waking hours over 7 days and signal when the care-recipient takes medications and engages in problematic behavior. Within 2 weeks the caregiver will be shown video clips of his or her FPV data as part of a nurse- delivered intervention focused on handling problem behaviors, then wear the FPV devices for another 7 days and receive brief telephone contact to reinforce the intervention, which we anticipate will reduce problem behaviors and caregiver burden. This pilot intervention study will be replicated with the second 20 dyads, though the nurse will access FPV data remotely and intervene by phone within 1 hour of problem events. FPV data gathered in these investigations will be used to develop algorithms for an innovative, automated tool for objective assessment and timely, tailored intervention aimed at improving caregiving quality and safety.